Bourbaki

Proposition 6

Tag 0COY, permanent. Cite it as [Bourbaki, Tag 0COY].

Commutative Algebra, chapter VIII, § 4. Séries de Hilbert-Samuel

Let A be a noetherian ring, M a finitely generated A-module and q an ideal of A such that M/qM is of finite length. Let F be a q-good filtration of M, and let gr(A)=⨁n≥0(qn/qn+1)\mathrm{gr}(A) = \bigoplus_{n \geq 0} (q^n/q^{n+1}) be the graded ring associated with A for the q-adic filtration.

Let (x1,...,xs)(x_1, ..., x_s) be a sequence of elements of A, (δ1,...,δs)(\delta_1, ..., \delta_s) a sequence of strictly positive integers such that xi∈qδix_i \in q^{\delta_i} for 1≤i≤s1 \leq i \leq s, and let ξi\xi_i be the class of xix_i in grδi(A)=qδi/qδi+1\mathrm{gr}_{\delta_i}(A) = q^{\delta_i}/q^{\delta_i+1}.

a) Endow the A-module M‾=M/(x1M+⋯+xsM)\overline{M} = M/(x_1M + \cdots + x_sM) with the q-good filtration F‾\overline{F} quotient of F. Then one has in Z((T))\mathbf{Z}((T)) the inequality

HM‾,F‾(s)≥(∏i=1s1−Tδi1−T)⋅HM,F. H_{\overline{M}, \overline{F}}^{(s)} \geq \left( \prod_{i=1}^s \frac{1 - T^{\delta_i}}{1 - T} \right) \cdot H_{M, F}.

b) For there to be equality in (22), it is necessary and sufficient that the sequence (ξ1,...,ξs)(\xi_1, ..., \xi_s) of elements of the ring gr(A)\mathrm{gr}(A) be completely secant for the module gr(M)=⨁n(Fn/Fn+1)\mathrm{gr}(M) = \bigoplus_{n} (F_n/F_{n+1}).

In that case, the canonical homomorphism of gr(M)/∑i=1sξi⋅gr(M)\mathrm{gr}(M)/\sum_{i=1}^s \xi_i \cdot \mathrm{gr}(M) into gr(M‾)=⨁(F‾n/F‾n+1)\mathrm{gr}(\overline{M}) = \bigoplus (\overline{F}_n/\overline{F}_{n+1}) is an isomorphism.

c) Assume the conditions of b) satisfied, and that each of the A-modules Mi=M/(x1M+⋯+xiM)M_i = M/(x_1M + \cdots + x_iM) (0≤i<s0 \leq i < s) is separated for the q-adic topology¹. Then the sequence (x1,...,xs)(x_1, ..., x_s) is completely secant for the A-module M.

When s=1s = 1, one has ⋂nFn=⋂nqnM\bigcap_n F_n = \bigcap_n q^nM and the sequence {ξ1}\{ \xi_1 \} is completely secant for gr(M)\mathrm{gr}(M) if and only if the homothety of ratio ξ1\xi_1 in gr(M)\mathrm{gr}(M) is injective. Prop. 6 then follows at once from lemma 7 applied to the homothety f=(x1)Mf = (x_1)_M in M.

Assume that one has s≥2s \geq 2 and argue by induction on s. The induction hypothesis applied to the A-module M1=M/x1MM_1 = M/x_1M endowed with the filtration G quotient of F, and to the sequence (x2,...,xs)(x_2, ..., x_s) yields the inequality

HM1,G(s−1)≥(∏i=2s1−Tδi1−T)⋅HM1,G; H_{M_1, G}^{(s-1)} \geq \left( \prod_{i=2}^s \frac{1 - T^{\delta_i}}{1 - T} \right) \cdot H_{M_1, G};

¹ This occurs in particular if q is contained in the radical of A (III, § 3, no. 3, prop. 6).

there is equality if and only if the sequence (ξ2,...,ξs)(\xi_2, ..., \xi_s) is completely secant for the gr(A)-module gr(M1)=⨁nGn/Gn+1gr(M_1) = \bigoplus_n G_n/G_{n+1}. Since the elements 1−Tδi1−T\frac{1 - T^{\delta_i}}{1 - T} of Z((T))\mathbf{Z}((T)) are positive, the case s=1s = 1 already treated and formula (23) yield the inequalities

HM,F(s)≥(∏i=2s1−Tδi1−T)⋅HM1,G(1)≥(∏i=1s1−Tδi1−T)⋅HM,F. H^{(s)}_{M,F} \geq \left( \prod_{i=2}^s \frac{1 - T^{\delta_i}}{1 - T} \right) \cdot H^{(1)}_{M_1,G} \geq \left( \prod_{i=1}^s \frac{1 - T^{\delta_i}}{1 - T} \right) \cdot H_{M,F}.

This proves a).

There can be equality in (22) only if one has simultaneously equality in (23) and the equality

HM1,G(1)=(1−Tδ11−T)⋅HM,F. H^{(1)}_{M_1,G} = \left( \frac{1 - T^{\delta_1}}{1 - T} \right) \cdot H_{M,F}.

This last relation signifies that {ξ1}\{\xi_1\} is completely secant for gr(M) and implies that the canonical homomorphism of gr(M)/ξ1⋅\xi_1\cdotgr(M) into gr(M_1) is an isomorphism. In other words, one has equality in (22) if and only if {ξ1}\{\xi_1\} is completely secant for gr(M) and {ξ2,...,ξs}\{\xi_2, ..., \xi_s\} completely secant for gr(M)/ξ1⋅\xi_1\cdotgr(M). This signifies that {ξ1,...,ξs}\{\xi_1, ..., \xi_s\} is completely secant for gr(M) (A, X, p. 160, Corollary 2). Thus the equivalence of the two conditions of b) has been proved. Suppose them satisfied; then, by the induction hypothesis, gr(M) is identified with gr(M1)/∑i=2sξi⋅gr(M1)gr(M_1)/\sum_{i=2}^s \xi_i \cdot gr(M_1); since moreover gr(M_1) is identified with gr(M)/ξ1⋅\xi_1\cdotgr(M), the last assertion of b) is therefore satisfied.

Now suppose that {ξ1,...,ξs}\{\xi_1, ..., \xi_s\} is completely secant for gr(M) and M_i separated for the q-adic topology (for 0≤i<s0 \leq i < s). From the preceding and the induction hypothesis, the sequence (x2,...,xs)(x_2, ..., x_s) is completely secant for M_1; since one has M1=M/x1MM_1 = M/x_1M and {x1}\{x_1\} is completely secant for M, the sequence (x1,x2,...,xs)(x_1, x_2, ..., x_s) is completely secant for M (A, X, p. 160, Theorem 1).

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Nothing.

Translated with machine help, not yet reviewed by a person.